Ghosting recognition method and device and computer storage medium

Abstract

The invention provides a ghosting recognition method and device and a computer storage medium. The ghosting recognition method comprises the steps that a plurality of target frames and the point cloud category of a laser radar point cloud are acquired; every two target frames are traversed, at least one target frame pair is matched and associated, and the target frame pair comprises a far-end target frame and a near-end target frame; extracting a far-end point cloud of the far-end target frame and a near-end point cloud of the near-end target frame based on the point cloud category; obtaining a far-end projection image of the far-end point cloud and a near-end projection image of the near-end point cloud; judging whether the pixel points of the far-end projection image are surrounded by the pixel points of the near-end projection image or not; and if yes, identifying the far-end target frame as a ghosting frame. Through the above mode, the method has high detection efficiency and high accuracy, reduces the number of data processing, and improves the calculation efficiency of a ghosting removing algorithm.

Description

Technical Field

[0001] The present application relates to the field of point cloud recognition technology, and in particular to a ghost recognition method, device and computer storage medium. Background Art

[0002] 3D target detection based on LiDAR can obtain accurate 3D position information and semantic information of the target, and has become an important part of intelligent driving technology. However, since LiDAR is an active imaging technology, it relies on the return signal of the laser to perceive the target, so the point cloud information it obtains will be disturbed by the environment to form LiDAR ghosts, causing false detection of the algorithm and posing a safety threat to the autonomous driving of intelligent vehicles.

[0003] In the prior art, only semantic information of the image is used to determine whether a ghost exists in the radar, which may result in missed detection and false detection of ghosts. Summary of the invention

[0004] The present application provides a ghost recognition method, device and computer storage medium.

[0005] To solve the above technical problems, the present application proposes a ghost recognition method, which includes: obtaining a number of target frames and point cloud categories of a lidar point cloud; traversing each pair of target frames, matching and associating at least one target frame pair, wherein the target frame pair includes a far-end target frame and a near-end target frame; extracting a far-end point cloud of the far-end target frame and a near-end point cloud of the near-end target frame based on the point cloud category; obtaining a far-end projection image of the far-end point cloud and a near-end projection image of the near-end point cloud; determining whether all pixels of the far-end projection image are surrounded by pixels of the near-end projection image; if so, identifying the far-end target frame as a ghost frame.

[0006] Among them, identifying the remote target frame as a ghost frame includes: identifying the remote target frame as a candidate ghost frame; obtaining the track information of the remote target frame; judging whether the number of tracking frames of the remote target frame is greater than or equal to a preset stability threshold based on the track information; if so, determining the candidate ghost frame as the ghost frame; if not, eliminating the candidate ghost frame.

[0007] Among them, traversing the target frames in pairs and matching and associating at least one target frame pair include: traversing the target frames in pairs, obtaining the first center point of the first target frame, and the second center point of the second target frame; obtaining a straight line between the first center point and the origin of the lidar point cloud; determining whether the distance from the second center point to the straight line is less than a preset distance threshold; if so, marking the first target frame and the second target frame as the target frame pair.

[0008] Among them, the target frame in the target frame pair that is farther from the origin of the laser radar point cloud is the far-end target frame, and the target frame that is closer to the origin of the laser radar point cloud is the near-end target frame.

[0009] Wherein, the acquiring of the far-end projection image of the far-end point cloud and the near-end projection image of the near-end point cloud includes: projecting the far-end point cloud onto a front view plane to form the far-end projection image; and projecting the near-end point cloud onto a front view plane to form the near-end projection image.

[0010] Among them, the determining whether all pixels of the far-end projection image are surrounded by pixels of the near-end projection image includes: selecting a pixel from the far-end projection image as a starting point for a depth-first search; recursively exploring the neighborhood of the starting point to determine whether all neighbors of the starting point are surrounded by pixels of the near-end projection image until all pixels of the far-end projection image are traversed.

[0011] Among them, the recursive exploration of the neighborhood of the starting point includes: before each recursive exploration, determining whether the current point meets the termination condition; if so, stopping the recursive exploration of the current point and obtaining the current neighborhood of the current point; if not, continuing the recursive exploration of the current point.

[0012] Among them, the judging whether the current point meets the termination condition includes: judging whether the current point exceeds the image boundary, wherein the image boundary is a circumscribed rectangle determined by the intersection of the pixel points of the far-end projection image and the pixel points of the near-end projection image; and / or judging whether the current point belongs to a pixel point of the near-end projection image; and / or judging whether the current point does not belong to a pixel point of the near-end projection image or a pixel point of the far-end projection image.

[0013] In order to solve the above technical problems, the present application proposes a ghost recognition device, which includes a memory and a processor coupled to the memory; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the above ghost recognition method.

[0014] In order to solve the above technical problems, the present application proposes a computer storage medium, which is used to store program data. When the program data is executed by a computer, it is used to implement the above ghost recognition method.

[0015] Different from the prior art, the beneficial effects of the present application are as follows: the ghost recognition device obtains several target frames and point cloud categories of the laser radar point cloud; traverses the target frames in pairs, matches and associates at least one target frame pair, wherein the target frame pair includes a far-end target frame and a near-end target frame; extracts the far-end point cloud of the far-end target frame and the near-end point cloud of the near-end target frame based on the point cloud category; obtains the far-end projection image of the far-end point cloud and the near-end projection image of the near-end point cloud; determines whether the pixels of the far-end projection image are all surrounded by the pixels of the near-end projection image; if so, identifies the far-end target frame as a ghost frame. Through the above method, it has higher detection efficiency and higher accuracy, reduces the amount of data processing, and improves the computational efficiency of the ghost removal algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a flowchart of the first embodiment of the ghost recognition method provided by the present application;

[0018] Figure 2 It is a schematic diagram of the overall process of the ghost recognition method provided by this application;

[0019] Figure 3 is a flow chart of a second embodiment of the ghost recognition method provided by the present application;

[0020] Figure 4 is a flowchart of a third embodiment of the ghost recognition method provided by the present application;

[0021] Figure 5 It is a structural schematic diagram of an embodiment of a ghost recognition device provided by the present application;

[0022] Figure 6 It is a structural diagram of an embodiment of a computer storage medium provided by the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] The ghost recognition method of the present application is applied to a ghost recognition device, wherein the ghost recognition device of the present application can be a server, or a system composed of a server and a local terminal. Accordingly, the various parts of the ghost recognition device, such as various units, sub-units, modules, and sub-modules, can all be set in the server, or can be set in the server and the local terminal respectively.

[0025] Furthermore, the above-mentioned server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules, such as software or software modules used to provide distributed servers, or it can be implemented as a single software or software module, which is not specifically limited here. In some possible implementations, the ghost recognition method of the embodiment of the present application can be implemented by a processor calling computer-readable instructions stored in a memory.

[0026] The principle of ghost points mainly involves the repeated refraction and reflection of light in the multiple lenses in the camera module. When taking pictures, multiple refractions and reflections occur when light passes through multiple lenses in the camera module. Specifically, when light passes through each lens, it refracts once on each surface and may then reflect between the lenses. This multiple refraction and reflection causes the light to form multiple image points on the sensor, one of which is the main image and the others are ghost points.

[0027] One way to remove ghosts in the prior art is to project the results of semantic segmentation based on images into the radar point cloud, and use the semantic information of the image to determine whether there are ghosts in the radar. This method may cause missed detection and false detection of ghosts. This method relies on the semantic segmentation results of the image. If the image is mis-segmented, it may cause incorrect judgment of ghosts. In addition, for distant targets and small targets, the joint calibration accuracy of the lidar and camera is required to be high. If the sensor position moves or the calibration is incorrect, it will also cause incorrect judgment of ghosts. This method has poor robustness.

[0028] Another way to remove ghosts in the prior art is to use the characteristics of the original laser pulse signal to evaluate whether the point cloud obtained by the lidar has ghosts. This method requires processing the original point cloud data obtained, which has high computational complexity and occupies more computing resources. In addition, this method relies on the laser pulse signal obtained by the sensor and cannot be used for point cloud data that has already been obtained.

[0029] The ghost determination method of the present application can complete the determination of ghosts using only radar sensors, without the need to connect other sensors, and has high robustness. The present application relies on the 3D position relationship between point clouds to comprehensively determine whether there is a ghost. The method has low computational complexity, is less time-consuming, and has a wide range of applications.

[0030] In order to solve the technical problems in the prior art, this application proposes a ghost recognition method, please refer to Figure 1-Figure 2 , Figure 1 It is a flowchart of the first embodiment of the ghost recognition method provided by the present application.

[0031] like Figure 1 As shown, the specific steps are as follows:

[0032] Step S11: Obtain several target boxes and point cloud categories of the laser radar point cloud.

[0033] In an embodiment of the present application, a ghost recognition device obtains a number of target detection frames, wherein the target detection frame can be a target frame of any target in any image, can be a continuous target frame of the same target in multiple frames of images, or can be a target frame of different targets in one frame of image, and the present application does not make any specific limitations.

[0034] The image may be acquired through continuous multiple frames or discontinuous video, and the video may be a real-time video or a stored video.

[0035] Specifically, the ghost recognition device obtains the results of 3D target detection and semantic segmentation based on the laser radar in the current frame. Through 3D target detection, the 3D target frame of the obstacle can be obtained, and the specific information includes category, position, target size and target orientation. Through semantic segmentation, the category of the point cloud can be obtained.

[0036] Step S12: traverse each pair of target frames and match and associate at least one target frame pair.

[0037] The target frame pair includes a far-end target frame and a near-end target frame.

[0038] Among them, the target frame in the target frame pair that is farther from the origin of the laser radar point cloud is the far-end target frame, and the target frame that is closer to the origin of the laser radar point cloud is the near-end target frame.

[0039] In a specific embodiment of the present application, a method for determining a target frame pair is provided, as follows: a ghost recognition device traverses each pair of target frames to obtain a first center point of a first target frame and a second center point of a second target frame; obtains a straight line between the first center point and the origin of the lidar point cloud; determines whether the distance from the second center point to the straight line is less than a preset distance threshold; if so, marks the first target frame and the second target frame as the target frame pair.

[0040] Specifically, the ghost recognition device matches and associates the target frame obtained in step S11, and selects the suspected ghost target and the corresponding reflective object based on the principle of ghost point generation.

[0041] The ghost recognition device records the target frame as T = T1 + T2 + ... + T n During the processing, two target boxes T are randomly selected A and T B , ensure that T A ≠T B , calculate the target box T B The vertical distance d from the center point B to the line OA BOA If the distance d BOA If it is smaller than the preset threshold ∈, mark the target box (T A ,T B ) is a pair of target frames, and the target frame farther from the origin is selected as the far end, and the one closer to the origin is selected as the near end. A ,T B ) are recorded in the result set P, and the process is repeated until all target frames are processed, and finally a target frame pair set P that meets the conditions is returned.

[0042] The embodiment of the present application also calculates the target frame T by calculating B The vertical distance to the line OA is used to select the target frame pair set and make judgments on the selected target frame pairs, which reduces the amount of data processing and improves the computational efficiency of the ghost removal algorithm.

[0043] Step S13: extracting the far-end point cloud of the far-end target frame and the near-end point cloud of the near-end target frame based on the point cloud category.

[0044] The ghost recognition device selects a pair of target boxes T from the set P A and T B , where T A is the far target box, T B The point cloud data inside the far and near target boxes are extracted and recorded as P A and P B .

[0045] Step S14: acquiring a far-end projection image of the far-end point cloud and a near-end projection image of the near-end point cloud.

[0046] In one embodiment of the present application, the ghost recognition device projects the far-end point cloud onto the front view plane to form the far-end projection image; and projects the near-end point cloud onto the front view plane to form the near-end projection image.

[0047] Specifically, the ghost recognition device will P A and P B They are projected onto the front view plane to form the corresponding pseudo image I and point cloud P A and P B The pixels formed are P IA and P IB .

[0048] Step S15: determining whether all pixels of the far-end projection image are surrounded by pixels of the near-end projection image.

[0049] If yes, execute step S16.

[0050] On the pseudo image I, the far-end target box T A Internal point cloud P A Whether the projected pixel point is enclosed by the near target box T B Internal point cloud P B If this condition is met, the far target box T can be preliminarily determined. A Possibly a ghost frame.

[0051] On the pseudo image, determine whether the pixel points obtained by the point cloud and projection in the far-end target frame are included in the point cloud P in the near-end target frame. B If this condition is met, it can be preliminarily determined that the far target box may be a ghost box.

[0052] To achieve encirclement judgment, the depth-first search (DFS) algorithm is used in the embodiment of the present application. Figure 3 , Figure 3 It is a flowchart of the second embodiment of the ghost recognition method provided by the present application.

[0053] like Figure 3 As shown, the specific steps are as follows:

[0054] Step S21: selecting a pixel point from the remote projection image as a starting point for a depth-first search.

[0055] The ghost recognition device initializes the data structure. First, the image boundary is defined, which is the pixel point set P IA and P IBThen create a Boolean matrix to record whether each pixel has been visited, and the initial value of the matrix is ​​"unvisited". Then, from the far target box T A The point set P in IA Select a pixel point P as the starting point of the depth-first search (DFS).

[0056] Step S22: recursively explore the neighborhood of the starting point to determine whether all the neighborhoods of the starting point are surrounded by the pixels of the near-end projection image, until all the pixels of the far-end projection image are traversed.

[0057] Specifically, the ghost recognition device performs a depth-first search (DFS) recursion, starting from the selected pixel point p, performs a depth-first search, recursively explores the neighborhood of the point, and verifies whether each neighborhood is occupied by P. IB Area enclosed. In the recursive process, first determine whether the termination condition is met.

[0058] Furthermore, in one embodiment of the present application, before each recursive exploration, it is determined whether the current point meets the termination condition; if so, the recursive exploration of the current point is stopped and the current neighborhood of the current point is obtained; if not, the recursive exploration of the current point is continued.

[0059] A specific method for determining whether the current point meets the termination condition is as follows: determining whether the current point exceeds an image boundary, wherein the image boundary is a bounding rectangle determined by the intersection of the pixel points of the far-end projection image and the pixel points of the near-end projection image; and / or determining whether the current point belongs to a pixel point of the near-end projection image; and / or determining whether the current point does not belong to a pixel point of the near-end projection image or a pixel point of the far-end projection image.

[0060] Specifically, the ghost recognition device performs an out-of-bounds check: if the current point (x, y) exceeds the pseudo image boundary, the recursion is terminated. At this time, "not surrounded" is returned, indicating that the point is connected to the external area.

[0061] The shadow recognition device performs a visited check: if the current point has been visited, it means that the point has been processed in the previous recursion, and the point is skipped to avoid repeated visits.

[0062] The shadow recognition device determines whether it belongs to P IB Region: If the current point (x, y) belongs to the near end point cloud projection area, that is, the point set P IB The recursion terminates. This means that the point is IB Connected, continue to expand to other neighborhoods.

[0063] Whether it belongs to the blank area: If the current point (x, y) belongs to the blank area, that is, it does not belong to the point set P IB The points in the set P do not belong to the point set P. IA , which means that the point may be connected to the outside and continue to expand to other neighborhoods.

[0064] If the termination condition is not met, continue to recursively process the current point. IA , mark the current point (x, y) as visited to prevent repeated processing. Then recursively explore the neighboring points and recursively check the four neighbors of the current point (upper, lower, left, and right). Set the step size of the search in the x direction and y direction to dx and dy. For each neighboring point (x+dx, y+dy), continue to perform DFS search to explore whether it is connected to the external area. If a neighboring point is found to be connected to the external area during the recursive process, it returns "not surrounded". If all neighbors of the current point are surrounded by P IB If it is completely surrounded, it returns "Enclosed".

[0065] Furthermore, the ghost recognition device traverses all P IA For the point in the far target box T A The set of all pixels P IA , perform encirclement judgment. For each point P = (x, y) in P IA In the process, DFS search is performed to determine whether the point is connected to the external area. If a point is found to be connected to the external area during the DFS process, that is, the recursive return is "not surrounded", it means that the point is not surrounded by P. IB Surrounded, the final result is "not surrounded", T A is a normal target, otherwise it indicates a far target box T A Possibly a ghost frame.

[0066] Through the above method, the ghost recognition device projects the point cloud of the target frame to the front view and uses the depth-first search method to determine whether the target is blocked, so as to determine whether the target frame is a ghost frame, which has higher detection efficiency and higher accuracy.

[0067] Step S16: Identify the far-end target frame as a ghost frame.

[0068] The ghost recognition device uses a 3D multi-target tracking algorithm to associate and match the detected target frame, record its track information, and comprehensively judge whether the target is a ghost target based on the track information. For details, please refer to Figure 4 , Figure 4 It is a flowchart of the third embodiment of the ghost recognition method provided by the present application.

[0069] like Figure 4 As shown, the specific steps are as follows:

[0070] Step S31: Identify the far-end target frame as a candidate ghost frame.

[0071] Specifically, the ghost recognition device recognizes the far-end target frame as a candidate ghost frame.

[0072] Step S32: Acquire the track information of the remote target frame.

[0073] The ghost recognition device detects multiple target frames in the scene through a three-dimensional sensor, such as a laser radar, and obtains the three-dimensional position information of the target. The multi-target tracking algorithm is applied to associate and match the target frame detected in the current frame with the target frame of the historical frame. This process uses distance, speed and other feature information to determine the continuity of the target. Let the target frame T a The track history life is L a For the successfully matched target frame, its track information is established, including the target's location information and state changes, and the historical life of each target is recorded.

[0074] Step S33: determining whether the number of tracking frames of the remote target frame is greater than or equal to a preset stability threshold based on the track information.

[0075] If yes, execute step S34, if no, execute step S35.

[0076] The ghost recognition device performs status evaluation based on the recorded historical life information. If the historical life of the target frame exceeds the preset number of frames, its tracking state is set to stable. If the historical life of the target frame is less than the preset number of frames or the track has not been started, it is determined to be unstable.

[0077] In an embodiment of the present application, the preset number of frames may be 3 frames or any other number of frames, and may be modified according to actual needs.

[0078] Step S34: determining the candidate ghost frame as the ghost frame.

[0079] Step S35: Eliminate the candidate ghost frames.

[0080] For the targets identified as possible ghost frames in the previous steps, check their track status. If the target's track status is unstable, it is confirmed as a ghost frame. For the targets identified as possible ghost frames, if their track status is judged to be unstable, they are finally judged as ghost frames and deleted in track management. Delete the targets judged as ghost frames in track management and mark the original point cloud in them as ghost point cloud.

[0081] Through the above method, the 3D multi-target tracking algorithm is used to track the trajectory of the target frame, and the target track information is combined to comprehensively judge whether the target frame is a ghost, thereby reducing the probability of mistaken deletion and improving the accuracy of ghost detection.

[0082] In order to implement the ghost recognition method of the above embodiment, the present application also provides a ghost recognition device. Figure 5 , Figure 5 It is a structural schematic diagram of an embodiment of a ghost recognition device provided in the present application.

[0083] like Figure 5 As shown, the ghost recognition device 600 of this embodiment includes a processor 61 , a memory 62 , an input / output device 63 , and a bus 64 .

[0084] The processor 61 , the memory 62 , and the input / output device 63 are respectively connected to the bus 64 . The memory 62 stores a computer program. The processor 61 is used to execute the computer program to implement the ghost recognition method of the above embodiment.

[0085] In this embodiment, the processor 61 may also be referred to as a CPU (Central Processing Unit). The processor 61 may be an integrated circuit chip having the ability to process signals. The processor 61 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The processor 61 may also be a GPU (Graphics Processing Unit), also known as a display core, a visual processor, or a display chip, which is a microprocessor that is specifically used for image computing on computers, workstations, game consoles, and some mobile devices (such as tablet computers, smart phones, etc.). The purpose of the GPU is to convert and drive the display information required by the computer system, and to provide a line scan signal to the display to control the correct display of the display. It is an important component that connects the display and the computer motherboard. As an important component of the computer host, the graphics card is responsible for outputting display graphics. The general-purpose processor may be a microprocessor or the processor 61 may also be any conventional processor, etc.

[0086] The present application also provides a computer storage medium, such as Figure 6 As shown, the computer storage medium 700 is used to store a computer program 71. When the computer program 71 is executed by a processor, it is used to implement the method described in the embodiment of the ghost recognition method of the present application.

[0087] The method involved in the embodiment of the ghost recognition method of the present application, when implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a device, such as a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0088] The above description is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A ghost recognition method based on laser radar, characterized in that: The ghost recognition method comprises: Get several target boxes and point cloud categories of the LiDAR point cloud; Traversing each pair of target frames, matching and associating at least one target frame pair, wherein the target frame pair includes a far-end target frame and a near-end target frame; Extracting a far-end point cloud of the far-end target frame and a near-end point cloud of the near-end target frame based on the point cloud category; Acquire a far-end projection image of the far-end point cloud and a near-end projection image of the near-end point cloud; Determining whether all pixels of the far-end projection image are surrounded by pixels of the near-end projection image; If so, the far-end target frame is identified as a ghost frame.

2. The ghost recognition method according to claim 1, characterized in that: The step of identifying the far-end target frame as a ghost frame includes: Identify the far-end target frame as a candidate ghost frame; Obtaining the track information of the remote target frame; Determine whether the tracking frame number of the remote target frame is greater than or equal to a preset stability threshold based on the track information; If so, determining the candidate ghost frame as the ghost frame; If not, the candidate ghost frame is removed.

3. The ghost recognition method according to claim 1, characterized in that: The traversing each pair of target frames and matching and associating at least one target frame pair includes: Traverse each pair of target frames to obtain the first center point of the first target frame and the second center point of the second target frame; Obtaining a straight line between the first center point and the origin of the laser radar point cloud; Determine whether the distance from the second center point to the straight line is less than a preset distance threshold; If so, mark the first target box and the second target box as the target box pair.

4. The ghost recognition method according to claim 1 or 3, characterized in that: The target frame in the target frame pair that is farther from the origin of the laser radar point cloud is the far-end target frame, and the target frame that is closer to the origin of the laser radar point cloud is the near-end target frame.

5. The ghost recognition method according to claim 1, characterized in that: The acquiring of the far-end projection image of the far-end point cloud and the near-end projection image of the near-end point cloud comprises: Projecting the remote point cloud onto a front view plane to form the remote projection image; The proximal end cloud is projected onto a front view plane to form the proximal end projection image.

6. The ghost recognition method according to claim 1, characterized in that: The determining whether all pixel points of the far-end projection image are surrounded by pixel points of the near-end projection image includes: Selecting a pixel point from the remote projection image as a starting point for a depth-first search; The neighborhood of the starting point is explored recursively to determine whether all the neighborhoods of the starting point are surrounded by the pixels of the near-end projection image, until all the pixels of the far-end projection image are traversed.

7. The ghost recognition method according to claim 6, characterized in that: The recursively exploring the neighborhood of the starting point includes: Before each recursive exploration, determine whether the current point meets the termination condition; If so, stop the recursive exploration of the current point and obtain the current neighborhood of the current point; If not, continue the recursive exploration of the current point.

8. The ghost recognition method according to claim 7, characterized in that: The determining whether the current point meets the termination condition includes: Determining whether the current point exceeds an image boundary, wherein the image boundary is a bounding rectangle determined by the intersection of pixel points of the far-end projection image and pixel points of the near-end projection image; and / or, determining whether the current point belongs to a pixel point of the near-end projection image; And / or, determining whether the current point does not belong to a pixel point of the near-end projection image or a pixel point of the far-end projection image.

9. A ghost recognition device, characterized in that: The ghost recognition device includes a memory and a processor coupled to the memory; The memory is used to store program data, and the processor is used to execute the program data to implement the ghost recognition method as described in any one of claims 1 to 8.

10. A computer storage medium, characterized in that: The computer storage medium is used to store program data, and when the program data is executed by a computer, it is used to implement the ghost recognition method according to any one of claims 1 to 8.

Images